INDEX

Explanations

references to battles involving good versus evil elements

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 uſed

-0.55

 AssemblyCulture

-0.54

]")]

-0.53

værende

-0.51

createCanvas

-0.50

complexContent

-0.50

DeleteBehavior

-0.49

antası

-0.47

antMatchers

-0.47

initializeApp

-0.46

POSITIVE LOGITS

 threats

0.60

 adversaries

0.59

 enemies

0.55

 foes

0.55

 hostile

0.55

 enemy

0.54

enemy

0.53

 pesky

0.50

 Threats

0.50

 rivals

0.50

Activations Density 0.084%

references to battles involving good versus evil elements

No Comments

No Known Activations